Propeller optimization by interactive genetic algorithms and machine learning

Abstract

Marine propeller design can be carried out with the aid of automated optimization, but experience shows that a such an approach has still been inferior to manual design in industrial scenarios. In this study, the automated propeller design optimization is evolved by integrating human–computer interaction as an intermediate step. An interactive optimization methodology, based on interactive genetic algorithms (IGAs), has been developed, where the blade designers systematically guide a genetic algorithm towards the objectives. The designers visualize and assess the shape of the blade cavitation and this evaluation is integrated in the optimization method. The IGA is further integrated with a support-vector machine model, in order to avoid user fatigue, IGA's main disadvantage. The results of the present study show that the IGA optimization searches solutions in a more targeted manner and eventually finds more non-dominated feasible designs that also show a good cavitation behaviour in agreement with designer preference.